From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934791AbZAOVD0 (ORCPT ); Thu, 15 Jan 2009 16:03:26 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S933402AbZAOUco (ORCPT ); Thu, 15 Jan 2009 15:32:44 -0500 Received: from tomts16-srv.bellnexxia.net ([209.226.175.4]:43839 "EHLO tomts16-srv.bellnexxia.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932354AbZAOUcm (ORCPT ); Thu, 15 Jan 2009 15:32:42 -0500 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: ApsEAIksb0lMQWt2/2dsb2JhbACBbM0nhXE Date: Thu, 15 Jan 2009 15:32:28 -0500 From: Mathieu Desnoyers To: Gaurav Singh Cc: ltt-dev@lists.casi.polymtl.ca, linux-kernel@vger.kernel.org Subject: Re: PMC support for ARMv6 Message-ID: <20090115203228.GA4284@Krystal> References: <3cc7a0df0901132148r1a3c0fa8q4c8d5912655eccc2@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <3cc7a0df0901132148r1a3c0fa8q4c8d5912655eccc2@mail.gmail.com> X-Editor: vi X-Info: http://krystal.dyndns.org:8080 X-Operating-System: Linux/2.6.21.3-grsec (i686) X-Uptime: 15:27:06 up 14 days, 20:25, 3 users, load average: 0.46, 0.52, 0.47 User-Agent: Mutt/1.5.16 (2007-06-11) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Gaurav Singh (gausinghnsit@gmail.com) wrote: > Hi Mathieu / all, > I have integrated PMC in ARMv6 primarily for 2 aims: > 1. CPI calculations - this trace_mark is triggered triggered after X no. of > clock cycles. The number of instructions executed and the no. of clock > cycles is noted. > 2. Cache Hit rate calculation - Here I have not used the PMC registers in > the ARM processor - rather have used L2 Cache controller registers to find > out this information - but this does not provide information about the L1 > cache hit/miss behaviour. > > Any other suggestions on what else can be done in PMC? > I'd like to see this trace_mark turned into a tracepoint probe. Having a periodical interrupt which saves this information is good, but I would also like to have a tracepoint probe that we could attach at important execution sites, like sched change, irq, softirq, nmi entry/exit where we would read the performance counter. > Also have ported LTTng to 2.6.24 for two Nomadik Boards - NDK15 and > NDK20. Doing some tests and will send it over to you guys soon for your > comments. > I'd be interested in merging your patchesfor 2.6.28 into the LTTng tree. Please keep us posted. Mathieu > Cheers! > Have a Nice Day! > Gaurav -- Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68